Article
First Results of the “Carbonaceous Aerosol in Rome
and Environs (CARE)” Experiment: Beyond Current
Standards for PM
10
Francesca Costabile1,* ID, Honey Alas2, Michaela Aufderheide3, Pasquale Avino4,5 ID,
Fulvio Amato6, Stefania Argentini1, Francesca Barnaba1 ID, Massimo Berico7, Vera Bernardoni 8 ID, Riccardo Biondi1, Giulia Calzolai9, Silvia Canepari10, Giampietro Casasanta1,
Spartaco Ciampichetti1, Alessandro Conidi1, Eugenia Cordelli11, Antonio Di Ianni1,12, Luca Di Liberto1, Maria Cristina Facchini13, Andrea Facci12, Daniele Frasca10 ID,
Stefania Gilardoni13, Maria Giuseppa Grollino7, Maurizio Gualtieri7 ID, Franco Lucarelli9, Antonella Malaguti7, Maurizio Manigrasso4, Mauro Montagnoli14, Silvia Nava9,
Elio Padoan5,15 ID, Cinzia Perrino14, Ettore Petralia7, Igor Petenko1, Xavier Querol6,
Giulia Simonetti10, Giovanna Tranfo4, Stefano Ubertini12, Gianluigi Valli8, Sara Valentini8, Roberta Vecchi8, Francesca Volpi13, Kay Weinhold2, Alfred Wiedensohler2, Gabriele Zanini7 and Gian Paolo Gobbi1
1 CNR-ISAC—Italian National Research Council, Institute of Atmospheric Science and Climate,
via Fosso del Cavaliere 100, 00133 Rome, Italy; s.argentini@isac.cnr.it (S.A.); f.barnaba@isac.cnr.it (F.B.); riccardo.biondi@artov.isac.cnr.it (R.B.); g.casasanta@isac.cnr.it (G.C.); s.ciampichetti@isac.cnr.it (S.C.); a.conidi@isac.cnr.it (A.C.); antonio.diianni@artov.isac.cnr.it (A.D.I.); l.diliberto@isac.cnr.it (L.D.L.); i.petenko@isac.cnr.it (I.P.); g.gobbi@isac.cnr.it (G.P.G.)
2 Leibniz Institute for Tropospheric Research, Permoserstrasse 15, 04318 Leipzig, Germany;
alas@tropos.de (H.A.); weinhold@tropos.de (K.W.); ali@tropos.de (A.W.)
3 Cultex Laboratories GmbH Feodor-Lynen-Straße 21, 04318 Hannover, Germany;
M.Aufderheide@cultex-laboratories.com
4 INAIL ex-ISPESL, via Urbana 167, 00184 Rome, Italy; p.avino@inail.it (P.A.); m.manigrasso@inail.it (M.M.);
g.tranfo@inail.it (G.T.)
5 Department of Agricultural, Environmental and Food Sciences, University of Molise, 86100 Campobasso,
Italy; elio.padoan@unito.it
6 Institute of Environmental Assessment and Water Research (IDÆA), Spanish National Research Council
(CSIC), 08034 Barcelona, Spain; fulvio.amato@idaea.csic.es (F.A.); xavier.querol@idaea.csic.es (X.Q.)
7 ENEA SSPT-MET-INAT, Via Martiri di Monte Sole 4, 40129 Bologna, Italy; massimo.berico@enea.it (M.B.);
maria.grollino@enea.it (M.G.G.); maurizio.gualtieri@enea.it (M.G.); antonella.malaguti@enea.it (A.M.); ettore.petralia@enea.it (E.P.); gabriele.zanini@enea.it (G.Z.)
8 Department of Physics, Università degli Studi di Milano and INFN-Milan, 20133 Milan, Italy;
vera.bernardoni@unimi.it (V.B.); gianluigi.valli@unimi.it (G.V.); sara.valentini@unimi.it (S.V.); roberta.vecchi@unimi.it (R.V.)
9 INFN, National Institute of Nuclear Physics, Florence, 50019 Sesto Fiorentino, Italy; calzolai@fi.infn.it (G.C.);
lucarelli@fi.infn.it (F.L.); nava@fi.infn.it (S.N.)
10 Department of Chemistry, “Sapienza” University, Rome, P.le A. Moro 5, 00185 Rome, Italy;
silvia.canepari@uniroma1.it (S.C.); daniele.frasca@uniroma1.it (D.F.); giulia.simonetti@uniroma1.it (G.S.)
11 ENEA SSPT-TECS-BIORISC Via Anguillarese, 00123 Rome, Italy; eugenia.cordelli@enea.it
12 DEIM—Industrial Engineering School, University of Tuscia, Largo dell’Università snc, 01100 Viterbo, Italy;
andrea.facci@unitus.it (A.F.); stefano.ubertini@unitus.it (S.U.)
13 CNR-ISAC—Italian National Research Council, Institute of Atmospheric Science and Climate,
via Gobetti 101, 40129 Bologna, Italy; mc.facchini@isac.cnr.it (M.C.F.); s.gilardoni@isac.cnr.it (S.G.); f.volpi@isac.cnr.it (F.V.)
14 CNR-IIA—Italian National Research Council, Institute of Atmospheric Pollution Research,
Monterotondo Stazione, via Salaria km 29300, CP10, 00015 Rome, Italy; montagnoli@iia.cnr.it (M.M.); perrino@iia.cnr.it (C.P.)
15 Dipartimento di Scienze Agrarie, Forestali e Alimentari (DISAFA), Università degli Studi di Torino,
Grugliasco, 10095 Torino, Italy
* Correspondence: f.costabile@isac.cnr.it; Tel.: +39-06-4993-4288
Received: 16 October 2017; Accepted: 6 December 2017; Published: 12 December 2017
Abstract:In February 2017 the “Carbonaceous Aerosol in Rome and Environs (CARE)” experiment was carried out in downtown Rome to address the following specific questions: what is the color, size, composition, and toxicity of the carbonaceous aerosol in the Mediterranean urban background area of Rome? The motivation of this experiment is the lack of understanding of what aerosol types are responsible for the severe risks to human health posed by particulate matter (PM) pollution, and how carbonaceous aerosols influence radiative balance. Physicochemical properties of the carbonaceous aerosol were characterised, and relevant toxicological variables assessed. The aerosol characterisation includes: (i) measurements with high time resolution (min to 1–2 h) at a fixed location of black carbon (eBC), elemental carbon (EC), organic carbon (OC), particle number size distribution (0.008–10 µm), major non refractory PM1 components, elemental composition,
wavelength-dependent optical properties, and atmospheric turbulence; (ii) 24-h measurements of PM10and PM2.5mass concentration, water soluble OC and brown carbon (BrC), and levoglucosan;
(iii) mobile measurements of eBC and size distribution around the study area, with computational fluid dynamics modeling; (iv) characterisation of road dust emissions and their EC and OC content. The toxicological assessment includes: (i) preliminary evaluation of the potential impact of ultrafine particles on lung epithelia cells (cultured at the air liquid interface and directly exposed to particles); (ii) assessment of the oxidative stress induced by carbonaceous aerosols; (iii) assessment of particle size dependent number doses deposited in different regions of the human body; (iv) PAHs biomonitoring (from the participants into the mobile measurements). The first experimental results of the CARE experiment are presented in this paper. The objective here is to provide baseline levels of carbonaceous aerosols for Rome, and to address future research directions. First, we found that BC and EC mass concentration in Rome are larger than those measured in similar urban areas across Europe (the urban background mass concentration of eBC in Rome in winter being on average 2.6±2.5 µg·m−3, mean eBC at the peak level hour being 5.2 (95% CI = 5.0–5.5) µg·m−3). Then, we discussed significant variations of carbonaceous aerosol properties occurring with time scales of minutes, and questioned on the data averaging period used in current air quality standard for PM10(24-h). Third, we showed
that the oxidative potential induced by aerosol depends on particle size and composition, the effects of toxicity being higher with lower mass concentrations and smaller particle size. Albeit this is a preliminary analysis, findings reinforce the need for an urgent update of existing air quality standards for PM10and PM2.5with regard to particle composition and size distribution, and data averaging
period. Our results reinforce existing concerns about the toxicity of carbonaceous aerosols, support the existing evidence indicating that particle size distribution and composition may play a role in the generation of this toxicity, and remark the need to consider a shorter averaging period (<1 h) in these new standards.
Keywords: carboanceous aerosol; black carbon; Mediterranean; Rome; brown carbon; optical absorption properties; aerosol health effects; high-time resolution; number size distribution; toxicology
1. Introduction
There is evidence that ambient particulate matter (PM10and PM2.5) pollution poses severe risks to
the human health [1–10]. This has led governments to adopt air quality standards for PM10and PM2.5.
PM10is a heterogeneous mix of solid and liquid particles of different sizes (from few nanometers to
10 µm) and different composition (including carbonaceous material, trace metals, crustal material, nitrates, sulfates, sea salt, ammonium). The specific aerosol types responsible for health effects, among
those constituing PM10, remain uncertain, and no safe level for the exposure to PM10and PM2.5has
been found [4]. This has sparked the debate over the need to update these standards.
The carbonaceous aerosol has been suspected to be more toxic than other PM10
constituents [2,4,7,11]. In their modeling exercise to assess the contribution of outdoor air pollution sources to premature mortality on a global scale, Lelieveld et al. [2] considered the carbonaceous PM2.5
as five times more toxic than inorganic particles. It is not clear if carbonaceous aerosol health effects are due either to its chemical composition (as elemental carbon particles or as carrier of other organic and inorganic chemicals) or to its physical properties (particle size, number and surface area) [1,3,4,7–10]. The carbonaceous aerosol is mainly found in submicron atmospheric aerosol particles, and most of these particles are in the ultrafine particle (UFP) size range (diameter less than 100 nm) [12]. Toxicology of UFPs is an emerging discipline because the size of these particles facilitates both adverse health effects in the lung and effects extending beyond the respiratory tract. Evidences have been found between short-term exposures to UFPs and the cardio-respiratory health (no matter what the particle composition is) [4]. Associations between ultrafine particles and the health of the central nervous system have also been reported. Inhaled ultrafine particles may translocate to the brain [13] where they can cause adverse impacts by inflammation and oxidative stress, which are common mechanisms similar to those acting in the lung [14,15]. Recent cross over studies on UFP and daily mortality in Europe, however, still show contrasting results and question the lack of a standardized protocol for UFP data collection [16] .
The carbonaceous aerosol in the atmosphere comes as a complex mixture of aerosol types. Major components are black carbon (BC), organic aerosol (OA) or organic carbon (OC), and brown carbon (BrC). These are always internally or externally mixed with other components. BC and BrC are carbonaceous aerosols in their chemical composition, but are named after their light absorption properties: BC because it looks blackish, BrC because it looks brownish. The size of these particles span from a few nanometers to a few micrometers. Relevant microphysics (size, mixing state, number) constitute one of the greatest uncertainties in both models and observations [17]. Importantly, no reference method has been developed yet for measuring the carbonaceous aerosol. In particular, there is no an accepted standard to measure BC [18]. Thermal-optical methods combined to chemical methods have traditionally been used to measure EC and OC, although drawing a clear border between organic macro molecules of OC and small clusters of possibly amorphous EC is challenging [18]. Spectral optical methods have traditionally separated BC from the bulk aerosol, while more recently BrC has been connected to the wavelength dependence of light absorption [12]. Recent development of single-particle instruments capable of detecting BC (i.e., the single-particle soot photometer, SP2) has provided a method to obtain number size distributions and mixing state of refractory BC (rBC), although the particle size range is very limited (80–300 nm). Significant steps forward in the OA characterisation have been made in recent years thanks to the use of field-deployable, high-resolution, time-of-flight aerosol mass spectrometers [19].
Carbonaceous aerosol levels in Europe are still uncertain, and certainly largely variable across the different regions [20–23]. Cavalli et al. [22] suggested that the atmospheric concentrations of PM10
and PM2.5 carbonaceous constituents increase when moving from Scandinavia to Central Europe
towards the Mediterranean. The Mediterranean basin, characterized by low cloudiness and high incoming solar radiation, is indeed an area of particular sensitivity as far as air pollution and climate change are concerned, e.g., [24]. Modeling results by Lelieveld et al. [2] show that the mortality linked to outdoor air pollution (mostly by PM2.5) in this region would definitely not be low, and that
Italy would rank 18 among the countries with premature mortality by PM2.5and O3related diseases.
As per Lelieveld et al. [2]’s results, land traffic would be one of the source categories responsible for the largest impact on mortality in the European Countries around the Mediterranean. In urban areas of Southern Europe the relative contribution to carbonaceous aerosol from vehicle exhaust could be more significant than that of other sources—e.g., biomass burning from domestic heating [21,23].
In this paper, we present first results of carbonaceous aerosol measurements carried out in February 2017 in Rome. The aim of this paper is to provide baseline levels of carbonaceous aerosols for the urban area of Rome reducing the assessment uncertainties of aerosol optical, chemical and microphysical properties, and to address future research and policy directions. Measurements were carried out in the framework of the experiment entitled “Carbonaceous Aerosol in Rome and Environs (CARE)”. As the present paper is the first of more detailed papers about the results of this experiment, here we briefly describe the general idea and methods of the whole project. The CARE experiment addresses the following specific question: what is the color, size, chemical identity, and toxicity of BC and BrC in the urban background of Rome? To this end, a robust dataset of optical–microphysical–chemical properties of the carbonaceous aerosol was collected, together with toxicological data to characterise relevant human health exposure. Different techniques for measuring carbonaceous aerosol properties were coupled with the aim to improve data reliability. Number size distribution (0.008–10 µm), composition (EC, OC, and major components of nor-refractory PM1), mass
concentration, and wavelength-dependent optical properties (BC and BrC) of the bulk aerosol were measured at a fixed location in the downtown Rome with high-time resolution (from 1-min to 2-h). 24-h measurements of PM10and PM2.5mass concentration, EC/OC, Water Soluble OC (WSOC), and
water soluble BrC (WSBrC) and levoglucosan were carried out at the same site. Mobile measurements were carried out to assess the spatial gradients of BC particles near urban roadways and building surfaces, and the connections between these gradients and the atmospheric turbulence will be assessed through Computational fluid dynamics modelling (CFD). Typical road dust loadings and emission factors due to traffic resuspension for PM10and EC/OC were quantified with the aim to obtain the
chemical profile of road dust emissions (only thoracic fraction) to apply in receptor modelling to quantify road dust contribution to PM, OC and EC concentrations. The toxicological assessment included the preliminary evaluation of the potential impact of ultrafine particles on lung epithelia performed by directly exposing BEAS-2B cells to air pollution. Cells were cultured at the air liquid interface and exposed to particles by using the Culltex RFS-1 module. Finally, the oxidative potential associated with carbonaceous aerosols was assessed (2-h time resolution), particle size dependent number doses deposited in different regions of the human body was modeled, and biomonitoring of polyaromatic hydrocarbons (PAHs) was performed.
2. Materials and Methods 2.1. Measurement Site
Measurements were carried out in Rome (Italy) in the middle of the Mediterranean sea (Figure1), from 27 January to 28 February 2017.
Measurements were carried out at an urban background site, located in the city center of Rome, in a garden (“San Sisto”). The garden is not open to the public. The site is situated in between three traffic roads which are 100–800 m away. In particular, the distance to the nearest heavily trafficked road is 115 m (Figure1).
Figure 1.Location of the measurement site in the Carbonaceous Aerosol in Rome and Environs (CARE) campaign (red square, label 1) located in the Mediterranean region (upper left panel), in the city center of Rome (upper right panel), between three major traffic roads (bottom left panel), and in a green area (bottom right panel)—note the different scales of the maps.
2.2. Equipment Deployment
Equipment used during the CARE 2017 experiment is summarized in Table1, including relevant sampling conditions (e.g., relative humidity, sampling head, time resolution). The concept here was to deploy different measurement techniques (e.g., different measurement principles, different time resolution, different measurement strategy), and inter-compare data. The final objective was to both identify components and reduce biases inevitably included in the measurement.
Table 1. Equipment deployed during the CARE experiment and relevant experimental setup (see Abbreviations list).
Instrument Variable Sampling Head RH Time Resolution On/Off Line
MAAP eBC PM10 <30% 1 min on-line
Aethalometer eBC, AAE PM10 <30% 1 min on-line
Nephelometer σs PM10 <30% 1 min on-line
Field Sunset EC, OC PM2.5 ambient 2 h on-line
ACSM OA, SO24−, NH+4, NO−3, Chl− PM1 <30% 30 min on-line
PAS 2000 PAH ambient 5 min on-line Streaker sampler Elemental composition,σa, AAE PM10, PM2.5 ambient 1 h off-line
Filter sampler σa, AAE, WSOC, WSBrC PM2.5 ambient 24 h off-line
MPSS PNSD (8 < dm< 700 nm) PM10 <30% 5 min on-line
APS PNSD (0.5 < da< 20 µm) PM10 <30% 5 min on-line
PILS Oxidative stress ambient 2 h off-line ALI Toxicity of lung cells PM1.5 ambient 24 h off-line
Ultrasonic anemometer WS, TKE, ambient 10 Hz on-line Meteorological station hWS, T, P, TSR, RH 1 s on-line PBL mixing monitor 1 h on-line
Mobile measurements are described in Section2.3.15. The fixed equipment was installed on board of two units parked into the San Sisto garden (the max distance between different instruments was less than 10 m). The ACSM, APS, AE33, MAAP, Nephelometer, SMPS, ALI system, streaker sampler and meteorological station were mounted on one van. The Sunset, PILS, Stability monitor were mounted
on a different van. The PM2.5sampler, ultrasonic anemometer, and PAS 2000 were mounted outside.
The APS, AE33, MAAP, Nephelometer, SMPS, were connected to the same sampling line, having inner diameter = 16 mm, lenght = 4 m, flowrate = 16 lpm, Re<2000. This sampling line was equipped with PM10sampling head and a (large TROPOS-made) nafion drier. The ACSM used a different line
equipped with PM1 sampling head and a (small) nafion drier. The ALI system used a very short
sampling line (less than 1 m) located very close (less than 1 m) from the aerosol line. Instruments are described in more detail in Section2.3.1.
2.3. Aerosol Measurements 2.3.1. Equivalent Black Carbon
The eBC mass concentration was obtained by:
• a Multi Angle Absorption Photometer (MAAP, Thermo Scientific™ (Waltham, MA, USA)), • a 7 wavelenght (370, 470, 520, 590, 660, 880 and 950 nm) aethalometer (model A33, Magee
scientific [25])
According to instrument manufacturer, the eBC mass concentration from AE-33 was obtained from measurements at λ = 880 nm. Note that the mass absorption coefficient used from the manufacturer to obtain this eBC value is 7.77 m2·g−1[25].
Only data in between the 1st and 99th percentiles were retained. 2.3.2. Aerosol Absorption Coefficients from Online Measurements
The absorption coefficient (σa) at 637 nm was obtained from MAAP measurements after the
following correction (Equation (1)):
σa(637) =eBC·QBC·1.05 (1)
where eBC is the equivalent mass concentration of BC reported by the instrument, QBC= 6.6 m2·g−1
is the specific absorption coefficient of BC used in the firmware of MAAP, and the factor 1.05 accounts for a wrong wavelength during calibration experiments when determining QBC= 6.6 m2·g−1[26].
This σawas used to calculate the SSA.
To calculate the AAE of the bulk aerosol at 470–660 nm (AAE470−660), the σafrom AE33 at 470 and
660 nm were obtained by converting BC data provided by the AE33 with Mass absorption cross-section values indicated by the manufacturer (14.54 and 10.35 m2·g−1, respectively [25]).
2.3.3. Absorption Coefficients from Filter-Based Multi-λ Polar Photometry
PM2.5samples collected with 24-h resolution and streaker samples having 1-h resolution were
measured by multi-λ polar photometry at the Physics Department of the University of Milan (Italy) in order to retrieve absorption coefficients at four different wavelengths. The PP-UniMI polar photometer—a benchtop instrument developed by the Milan research group—was extensively described in [27,28] . Briefly, the instrument is based on the measurement on the scattering plane of the light transmitted and scattered in the forward and back hemispheres by unloaded and loaded samples using a photodiode mounted on a rotating arm. Data reduction aiming at the determination of the sample absorbance is performed according to [29] and literature cited therein. Currently, PP-UNIMI allows performing 4-λ measurements (780, 635, 532, and 405 nm) on aerosol collected on different substrates, including high-time resolved samples obtained using a streaker sampler. The set-up of the instrument was validated against independent measurements carried out using a Multi-Angle Absorption Photometer (MAAP) for what concerns the red-light results, considering possible artefact effects shown in [27].
2.3.4. Scattering Coefficient
The dry scattering coefficient (σs(λ)) at 450, 525, and 635 nm was measured online with 1 min time resolution by an integrating nephelometer (Model Aurora 3000, Ecotech, Australia). Nephelometer data were corrected for truncation (Anderson and Ogren, 1998; Bond, 2001; Müller et al., 2011). The scattering error after the truncation error correction is δ(σs)
σs =0.02 [12]. The minimum detection
limit of the nephelometer is 0.3 Mm−1, with calibration tolerance of±4 Mm−1and measurement range 0–2000 Mm−1.
2.3.5. OC/EC
The EC/OC mass concentration with 2-h time resolution was obtained by a Sunset Field Thermal-Optical Analyser (Sunset Laboratory Inc., Johannesburg, South Africa). Briefly, this instrument collects PM2.5on a quartz fiber filter and automatically analyses it at the end of each sampling period.
The instrument inlet is equipped with a cyclone (cut point 2.5 µm) and an organic denuder. In this campaign, a time resolution of 2 h (105 min of sampling followed by 15 m of analysis) was chosen as a compromise to get an adequate time resolution (comparable with that of other instruments used in this project) and a sufficient amount of collected sample mass (to maintain a good accuracy in the EC-OC quantification). The instrument was calibrated by sucrose standards and the NIOSH protocol [30] was used for thermal analysis. Further, EC and OC were determined on 24-h quartz fiber filters by thermo-optical analysis with an offline OCEC Carbon Aerosol Analyser (Sunset Laboratory Inc., Johannesburg, South Africa) by applying the NIOSH-QUARTZ temperature protocol.
2.3.6. PM2.5Sampler
Quartz-fiber filters were sampled with a PM2.5sampler on a daily basis. Filters were first analysed
for multi-wavelength optical absorption, and then cut and analysed for levoglucosan (1/4), water soluble BrC (3/8), and water soluble OC (3/8). This is described in the following sections.
2.3.7. Water Soluble OC
Water-soluble organic carbon (WSOC) was analysed on 24-h quartz fiber filters by TOC-VCSH (Shimadzu, Kyoto, Japan) by using the NPOC (non-purgeable organic carbon) procedure [31]. 2.3.8. Water Soluble BrC
Light absorption spectra of water soluble carbon were measured off-line. Two 6 mm punches of quartz PM2.5 filters were extracted in 5 mL of ultrapure milli-Q water by 30 min ultrasonication.
The extracts were filtered with 0.45 µm cutoff PTFE syringe filters and analyzed, immediately after extraction, in parallel with blank filter samples. The UV-visible spectra were recorded with a TIDAS spectrometer coupled with a 0.5-m path length Liquid Waveguide Capillary Column (LWCC). The absorption coefficient of water soluble brown carbon (BrC) was calculated as the absorption at 365 nm, drift-corrected, and normalized by the volume of sampled air, according to the equation:
Abs365= (A365−A700) ·
Ve
0.5·Va
·ln(10) (2)
A365 and A700 are the measured absorption at 365 and 700 nm, respectively, Veis the extraction
volume, Vsis the air sampling volume, 0.5 m is the optical path length. The wavelength of 365 nm is
selected in order to reduce interference from inorganic species (i.e., nitrate) dissolved into water, and is comparable to previous literature studies [32,33]
2.3.9. Levoglucosan
For the determination of levoglucosan, 24-h quartz fiber filters were extracted in de-ionized water and the solution analysed by high-performance anion-exchange chromatography with
pulsed amperometric detection (HPAEC-PAD), using a DC ICS-3000 oven, a GP40 gradient pump, a CarboPac™ PA10 analytical and guard column and a Dionex ED50/ED50A electrochemical cell. The same analytical method was used for the analysis of the solutions collected by the particle-into-liquid sampler (Section2.5.4).
2.3.10. Elemental Composition by Streaker
Measurements for the determination of PM2.5 elemental composition with hourly resolution
were performed at the INFN LABEC laboratory by means of Particle Induced X-ray Emission (PIXE) analysis. This technique is based on the detection and analysis of the X-rays emitted by the sample after excitation by an accelerated particle beam. PIXE allows the quantification of more than 20 elements with atomic number (Z)>10 in a single measurement lasting only some tenths of seconds. This is a multi-elemental technique. With our set-up we simultaneously detect the X-rays of all the elements with Z>10. In particular, we analysed the samples for the concentrations of: Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Cu, Zn, As, Se, Br, Rb, Sr, Zr, Mo, Ba, Pb. Nevertheless, some elements (e.g., Rb, Zr, Mo. . . ) were below detection limits in most cases.
Measurements were performed on samples collected with a streaker sampler [34] exploiting the external beam set-up fully dedicated to aerosol analysis available at LABEC. Details on the technique and the set-up, included the X-ray detection system, are given elsewhere [34,35]; here we briefly recall that the streak of collected particles is analysed point-by-point using a proton beam collimated in order to get a spot corresponding to 1-h sampling. Every point was measured for 60 s using a 300 nA beam; elemental concentrations were obtained via calibration relative to thin reference standards.
2.3.11. Non-Refractory PM1Chemical Components
Major Non-refractory PM1chemical components were measured using an Aerodyne Aerosol
Chemical Speciation Monitor (ACSM). The ACSM is a mass spectrometer capable of analysing in real-time, with a temporal resolution of 30 min, the non-refractory at 600 °C component of the fine (<1 µm) particulate (NR-PM1) [36]. The instrument analyses the mass spectrum of NR-PM1 by
acquiring 100 different macromolecules (associated with pre-fixed mass-to-charge, m/z), providing the measurement of the main organic and inorganic chemical components: Organic matter (OA), sulfate (SO2−4 ), ammonium (NH+4), nitrate (NO−3), Chloride (Cl−). For these technical features the ACSM allows, with respect to conventional measurement systems, (i) to have more information on the activity of the emission sources (variations within the daily cycle); (ii) to perform measurement campaigns also for limited periods; and (iii) to have a sufficient amount of data for application of multivariate statistical techniques that can contribute to the identification and quantification of the emission sources. The set of m/z provides the mass spectra for speciation and quantification of the particulate air pollution main components; combinations of the obtained spectra’s time series can give information, after processing with multivariate analysis (Positive Matrix Factorization, PMF), to identify eventual chemically distinct groups of pollutants. The PMF analysis on the data of the organics permits the pre-identification of a number of factors (emission profiles) [37].
2.3.12. Polycyclic Aromatic Hydrocarbons
The EcoChem Photoelectric Aerosol Sensor (PAS) 2000 (PAS 2000 CE, EcoChem Analytics, Texas, and Matter Engineering AG, Switzerland) is a real-time monitor that measures the surface-associated total polycyclic aromatic hydrocarbons (PAHs) concentration . The PAS 2000 works on the principle of the photoionization of the particle-bound PAHs: the fine particles on whose surface PAHs are adsorbed are ionized by UV radiation, the charged particles are collected on a filter element and the resulting piezoelectric current is measured and proportionally related to the particle-bound PAHs. The logged data in 5 min intervals (expressed as total PAH concentration in µg·m−3), were averaged over 30 min intervals for correlation with ACSM data.
2.3.13. Particle Number Size Distribution
The Particle number size distribution (PNSD) was measured by combining a Mobility Particle Size Spectrometer (TROPOS SMPS) equipped with a butanol-based condensation particle counter (CPC, TSI model 3772) and a commercial aerodynamic particle sizer (APS, TSI). Particles from 8 to 800 nm of electrical mobility diameter (dm) were sized and counted by the SMPS; particles from 0.5 to
20 µm of aerodynamic diameter (da) were sized and counted by the APS.
SMPS data were corrected for penetration errors through the sampling line (TROPOS-made software), penetration efficiency due to diffusion losses (calculated according to [38]) being higher than 98.92% for particles bigger than 15 nm.
APS aerodynamic diameters (da) were converted to electrical mobility diameters (dm) as
(Equation (3) [38]):
dm=da·r χ
·ρ0
ρp (3)
where ρ0is the reference density (1 g·cm−3); ρp is the particle density; and χ is the shape factor.
We assumed:
• size dependent effective particle density (ρp) continuously varying from 1.6 to 2 g·cm−3in the
APS aerodynamic particle size range (cf. Figure S24 of the Supplementary materials); • spherical particles (χ = 1);
• electrical mobility diameters (dm) representing the true particle diameter;
• Cunningham slip correction factor neglected in the APS size range.
To merge APS and SMPS, we used size distribution data expressed as dlog(ddN
m) similar to [39]
(the dlog(ddN
m)distribution does not need a vertical shift, whereas the dN distribution does). The dm-based
size distribution from APS ((dlog(ddN
m))APS) was calculated as (Equation (4)):
( dN dlog(dm) )APS = ( dN dlog(da) )APS· dlog(da ) dlog(dm) (4)
Then, the(dlog(ddN
m))APSwere merged to relevant SMPS data ((
dN
dlog(dm))SMPS). The(
dN
dlog(dm))SMPS
and(dlog(ddN
m))APSoverlapped for dmfrom 475 to 830 nm (in fact, the first two channels of the APS were
not used for the fitting procedure because of their unreliable counting and thus the overlapping size range was smaller). The(dlog(ddN
m))SMPSwere fitted with a power-law (Junge size distribution) function
in this overlap size range (Equation (5)):
( dN dlog(dm)
)f it =C·d−αm (5)
An iterative procedure was used to estimate C and α. A first guess-value is chosen and then varied based on the minimisation of the relative square difference between fitted and measured number size distributions at the two extreme particle diameters (Equations (6) and (7)):
( dN dlog(dm) f it− dN dlog(dm))APS dlog(dm)f it )2 (6) ( dN dlog(dm) f it− dN dlog(dm))SMPS dlog(dm)f it )2 (7)
2.3.14. PMx
Daily PM10 and PM2.5 data were measured at a nearby urban background station (Arenula,
3 km from the CARE site) from the local environemntal agency (ARPA Lazio). Also, PM1data were
reconstructed from the BC and the NR-PM1data.
In addition, PMxdata with 5 min resolution were reconstructed from the particle size distributions
data measured by SMPS and APS, after the fitting procedure described in Section2.3.13. A size dependent effective particle density was used (Figure S24 of the supplementary materials).
2.3.15. Mobile eBC Measurements
To determine the spatial variability of eBC mass concentration, parallel mobile measurements were performed using two TROPOS aerosol backpacks that measure concentrations of pollutants at a microscale level. Together with a GPS unit, data logger, and power supply, high time resolution, portable instruments that measure eBC (AE51) and particulate matter concentrations are placed inside a backpack, which can be carried by a single person.
Measurements or “runs” were done along a ∼9 km fixed route that includes different microenvironments to simulate different exposure scenarios. This route includes a 30-min stay at the fixed urban background station to ensure the data quality by comparing with the reference instruments for every run. Runs were performed three times a day (morning rush hour, afternoon, and evening) including weekends.
The data from the mobile measurements is the average of the two mobile instruments (AE51) running parallel to each other with a unit-to-unit variability of 0.068 µg · m−3. The data from the reference fixed instrument were taken from MAAP at the fixed station representin the urban background concentration of eBC. MAAP measurements showed good agreement with the mobile instruments during the 30-min inter-comparison periods.
2.3.16. Road Dust Sampling
RD10 was collected at 10 sites around the S. Sisto area, around the CARE monitoring site, by means of a field resuspension sampler, consisting of a deposition chamber, an elutriation filter, where particles>10 µm are separated, and a filter holder, where RD10 is collected [40]. Sites were chosen to represent highly trafficked sites and secondary residential streets. At each site, samples were collected following previous studies protocol [40,41]. Two filters were collected for each site, on the rightmost active lane. In some cases, sampling was performed on the curbside, therefore used only for the chemical profile. Before sampling, quartz fiber filters (Pall, 47 mm) were dried and conditioned before weighing; after sampling, filters were brought back to the laboratory for gravimetrical and chemical analyses [42]. Half of the sample was dissolved in acid (as in [42]) for the determination of major and trace elements by ICP-AES and ICP-MS, while another fraction was used for the determination of Organic and Elemental Carbon (OC and EC) by means of a Sunset Lab. ECOC analayzer using the EUSSAR2 protocol [43]. Pavement texture was analyzed by photographic analysis and by the Mean Texture Depth (MTD) analysis (ASTM E 965, 2015). Close-up photos of pavement surfaces were used to estimate average size of aggregates. Since aggregates are embedded in the asphalt binder, the surface macrostructure was also estimated by means of the MTD analysis [44]. The combination of the two analyses allowed obtaining the corrected aggregate median (CAM), according to the formula (Equation (8)):
CAMi=MTDi×
(AggregateMedian)i
(AggregateMean)i
(8) where the mean and the median of the aggregates correspond to the values of the horizontal size of the aggregates observed in the photographic analysis.
2.4. Meteorological Measurements
Meteorological variables (temperature, relative humidity, pressure, wind) were measured by a standard meteo variables (Lufft weather station) with 1 min time resolution. As turbulence is a very important phenomenon affecting atmospheric processes near the surface, high frequency measurements of the three wind components u, v, w, and virtual temperature Tvwere made with a
Metek uSonic-3 Scientific thermo-anemometer (sampling frequency 10 Hz) installed 3.5 m a.g.l. From these measurements we computed the wind speed and direction as well as the fluctuations u0, v0, w0, and Tv0 with respect to the 1-h linearly detrended mean wind components ( ¯u, ¯v, ¯w) and virtual
temperature ¯Tv(u0= u− ¯u, v0= v−¯v, w0= w−w, T¯ v0= Tv−Tv). We estimated the kinematic heat flux
(w0T0
v) and Turbulent Kinetic Energy (TKE) per unit mass (TKE/m = 12(u2+v2+w2)). The w0Tv0 gives
a measure of the thermal mixing capabilities of the atmosphere. The TKE represents the intensity of turbulence produced by fluid shear, friction or buoyancy, or through external forcing at low-frequency eddy scales. As both the w0T0
v and TKE vary significantly in time and in space, they need to be
monitored during the all day.
Interesting additional information about the mixing properties of the lower atmosphere can be obtained by monitoring natural radioactivity due to Radon progeny [45–47]. During CARE, natural radioactivity was continuously measured on a 1-h basis. Natural radioactivity was measured by means of an automated monitor of Radon progeny (PBL Mixing Monitor, FAI Instruments, Fonte Nuova, Rome, Italy). The instruments detects natural radioactivity due to the decay products of Radon, a gas emitted from the ground at a rate that can be assumed to be constant on our observation scales (a few weeks, some kilometers). Radon undergoes radioactive decay only, and the concentration of its short-life decay products matches the time variations in the concentration of unreactive atmospheric pollutants emitted with a constant rate. Thus, the study of the time pattern of natural radioactivity allows an easy uncoupling of pollutants concentration variations that are due to changes in their emission/transformation rates from variations that are due to changes in the mixing properties of the lower atmosphere.
2.5. Toxicological Data
2.5.1. Air Liquid Interface (ALI) System
The human bronchial epithelial cell line BEAS-2B (ECACC, Salisbury, UK) were maintained in LHC-9 medium at 37 °C with 5% of CO2, split every three days. 72 h before exposure BEAS-2B were
seeded on collagen coated transwell inserts (Corning Transwell®, Corning Inc., New York, NY, USA) at a density of 15×103cells/insert. Then, the transwell were transferred into the CULTEX® RFS Compact module (CULTEX Laboratories GmbH, Hannover, Germany) and the basal side of each chamber was filled with 4 mL of LHC-9 medium. The module was covered with a plastic seal and taken to the mobile lab for direct cell exposure. Cells were exposed at ALI to native atmosphere (insert at position 2, 4 and 6 of CULTEX® RFS Compact) or to filtered air (negative controls at position 1, 3 and 5) for subsequent 24 h. Ambient air was sampled at 1.50 L/min, particulate matter bigger than 1.5 µm was cut by a cyclone. 5 cm3·min−1out of the 1.50 L/min sampled were used to expose each insert according to the geometry of the radial RFS-1 module. The number and mass of theoretical maximal deposited particles was than calculated according to [48].
2.5.2. Aerosol Dosimetry
Aerosol doses deposited into the human respiratory system have been estimated using the Multiple-Path Particle model Dosimetry (MPPD v3.01, ARA 2015, ARA, Arlington, VA, USA, [49]). The 60th percentile human stochastic lung was considered along with the following settings: (i) a uniformly expanding flow; (ii) an upright body orientation; and (iii) nasal breathing with a 0.5 inspiratory fraction and no pause fraction. Moreover, the following parameters were used for a Caucasian adult male under light work physical activity, based on the ICRP report [50]: (i) a functional
residual capacity (FRC) of 3300 mL; (ii) an upper respiratory tract (URT) volume equal to 50 mL; (iii) a 20 min−1breathing frequency; and (iv) an air volume inhaled during a single breath (tidal volume, VT) of 1.25 L.
2.5.3. PAHs Biomonitoring
Two of the researchers working on the project site, a man and a woman, collected urine samples for 15 consecutive days. Simultaneous quantitation of five urinary monohydroxylated metabolites of four polycyclic aromatic hydrocarbons (PAHs) was carried out using an HPLC-MS/MS analytical method: 1-hydroxypyrene (1-OHPy), 1-hydroxynaphthalene (1-OHNAP), 2-hydroxynaphthalene (2-OHNAP), 3-hydroxybenzo[a]pyrene (3-OHBaPy), 6-hydroxynitropyrene (6-OHNPy) [51]. Results were normalized for the creatininuria.
2.5.4. Oxidative Potential
A particle-into-liquid sampler (PILS) was used for on-line measurement of the oxidative potential (OP). The same system was used for levoglucosan measurements. The air flow is denuded from gaseous species, while aerosol particles are grown in a saturated water vapor chamber to form droplets. These are collected by inertial impact on a collection plate, washed by deionized water. The resulting solution was collected every 2 h for subsequent analyses [47].
The OP was determined by the 2070-dichlorofluorescein (DCFH) essay [52]. The OP is considered as a proxy of the ability of PMxto generate reactive oxygen species (ROS, or free radicals) and is defined
as the capacity of PMxto oxidize target molecules. Different assays refer to different target molecules
and are expected to lead to different OP values. In this work we used the 20,70-dichlorofluorescein (DCFH) assay. This assay was formerly developed for the in vitro determination of ROS in biological cells (e.g., [53]). In recent years, it has been adapted and applied as an acellular method (e.g., [54]). In this assay, the non-fluorescent DCFH is oxidized to the fluorescent dichlorofluorescein (DCF) in the presence of horseradish peroxidase (HRP).
3. Results
Here we present preliminary results of the CARE experiment. First, an overview of aerosol and meteorology data is given in Section3.1. Then, aerosol characteristics are presented in Section3.2. Third, toxicological results are presented in Section3.3. These results are discussed in Secion4. 3.1. Overview of Aerosol and Meteorology Data
Figure2shows the time series of the major carbonaceous aerosol features ( mass concentration of eBC and non-refractory PM1components, spectral aerosol light absorption, and particle number
size distribution, described in detail in the following subsections), with relevant meteorological and micro-meteorological parameters (wind speed and direction, friction velocity and Turbulent Kinetic Energy (TKE), temperature and sensible heat flux (w0Tv0)). A statistical summary of these data is
presented in Table2. The Heat Flux (w0T0
v) shows thermal mixing capabilities of the atmosphere, whereas the TKE the
intensity of turbulence produced by fluid shear, friction or buoyancy, or through external forcing at low-frequency eddy scales (Section2.4). Similar information on the mixing properties of the lower atmosphere were obtained by measuring the natural radioactivity due to Radon progeny (Section2.4, Figures S23 and S24 of the Supplementary materials).
The presence and altitude of Saharan dust plumes have been determined on the basis of polarization-sensitive Lidar-ceilometer (PLC) observations and verified against model forecasts (e.g.,www.diapason-life.eu). The relevant near-real time atmospheric profiling was made by means of the PLC system located in central Rome (data available athttp://www.alice-net.eu/, “downtown Rome” site). In these measurements, suspended mineral dust is revealed by a sharp increase in the depolarization signal, i.e., the cross over the parallel polarized backscattered signals (an example is
Figure S27 of the Supplementary materials). During CARE, Saharan dust layers were observed in the periods 3–6, 8–10, 18 and 24–28 of February 2017. Dust was observed to reach the ground in the last two periods only.
Figure 2.Time series of (a) equivalent black carbon (eBC) mass concentration and absorption Ångström Exponent (color); (b) mass concentration of major components of non refractory PM1; (c) particle number
size distribution; (d) wind speed and direction; (e) frictional velocity (u∗) and Turbulent Kinetic Energy
(TKE); (f) temperature and heat flux at the S. Sisto site in Rome from 1 to 28 of February 2017.
During the first days of the campaign (days 1–6), aerosol concentrations were lower due to bad weather conditions (Figure2): strong winds from E-SE, high values of the momentum fluxes and TKE, low Radon concentrations (Figure S24 of the supplementary materials) and some showers (Figure S28 of the supplementary materials). Bad weather conditions also occurred on February 18–19 and 24–26 (i.e., days of Saharan dust advection). During the central period of the campaign (days 10–24, excluding the above mentioned days 18–19) the weather conditions were fairly stable. These conditions favoured
the increase of aerosol mass concentrations (Figure2). The typical local winter circulation prevailed, the peak of the wind speed (mostly below 2 m·s−1at 12 LT) being due to the inland penetration of weak sea breezes.
Table 2. Statistical summary of variables acquired from 16:00 2 February to 10:00 28 February 2017. Data are mean, median, standard deviation, minimum and maximum, 95% confidence interval (lower–upper), number of datapoints (#), and time resolution of data used to calculate the statistics (TR).
Variables Mean Median st.dev. Min–Max 95 % CI [mean] # TR
eBC (µg · m−3) MAAP 2.6 1.7 2.5 0.2–14.4 2.3–2.8 35,256 1 min MAAP 2.7 1.6 2.6 0.5–9.9 2.5–2.9 609 1 h MAAP 2.5 1.7 2.1 0.3–13.9 2.3–2.8 287 2 h AETH 2.9 1.9 2.8 0.3–16.0 2.9–3.0 35,839 1 min EC (µg · m−3) 2.1 1.9 1.4 0.3–10.7 1.9–2.3 302 2 h 2.2 2.1 1.1 0.6–4.6 1.7–2.6 26 24 h MAC (m2· g−1) 637 (MAAP,Sunset) 8.6 8.4 0.9 6.7–12.0 8.5–8.7 297 2 h OC (µg · m−3) ACSM 7.8 5.0 6.8 0.01–32.5 7.4–8.2 1135 30 min sunset 5.6 4.3 4.0 0.7–18.1 5.2–6.1 308 2 h NR-PM1(µg · m−3, ACSM)
total mass 13 9 10 0–50 13–14 1135 30 min NO3− 2.75 1.41 2.86 0.07–14.3 2.58–2.92 1135 30 min NH+4 1.60 1.41 1.15 0.00–6.54 1.53–1.65 1004 30 min SO−4 1.30 1.15 0.90 0.00–3.72 1.25–1.36 1122 30 min Cl− 0.17 0.09 0.32 0.00–6.29 0.15–0.19 942 30 min PM1(µg · m−3) ACSM+MAAP 16 11 12 2–54 14–17 287 2 h SMPS+APS 16 11 11 2–47 14–17 290 2 h PM2.5(µg · m−3) Beta 17 16 7 6–32 14–20 25 24 h SMPS + APS 18 15 11 3–51 17–20 279 2 h PM10(µg · m−3) Beta 26 27 9 9–45 22–30 24 24 h SMPS+APS 24 23 14 2–67 22–26 280 2 h BC/PM10 0.10 0.10 0.04 0.02–0.25 0.09–0.11 274 BC/PM2.5 0.13 0.12 0.06 0.03–0.41 0.13–0.14 274 BC/PM1 0.15 0.15 0.06 0.07–0.50 0.15–0.17 287 PAHs (ng · m−3) 24 11 33 0–252 21–26 951 Ntot·103(cm−3) 12.3 11.0 7.7 0.1–40.9 12.1–12.5 7412 5 min Ntot·103(cm−3) 13.0 11.5 6.9 2.9–38.6 12.3–12.9 299 2 h dmed(S)(nm) 81 86 23 10–115 80–81 7483 5 min σap(Mm−1) 405 (streaker) 25.1 15.3 26.3 5.9–180.4 23.0–27.1 628 1 h 532 (streaker) 19.2 11.4 20.7 2.5–147.2 17.6–20.86 628 1 h 635 (streaker) 16.24 9.74 17.84 3.4–123.26 14.86–17.66 628 1 h 637 nm (MAAP) 19.7 11.57 20.97 0.3–202.6 18.8–19.7 7300 5 min 780 (streaker) 13.5 7.2 15.3 3.4–104.0 12.3–14.7 628 1 h σsp(Mm−1) 520 nm (Neph) 77.7 61.1 53.2 10.3–262.9 35,727 1 min 635 nm (Neph) 62.2 52.6 40.7 9.2–204.5 35,727 1 min SSA
637 nm (Neph + MAAP) 0.778 0.796 0.105 0.278–0.969 0.776–0.781 7196 5 min AAE 450–660 (AETH) 1.41 1.38 0.21 0.99–2.03 1.40–1.41 35,935 1 min WS (m · s−1) 0.9 0.6 0.8 0.1–4.5 0.90–0.96 4032 1 min TKE (m2· s−2) 0.34 0.11 0.49 0.00–3.39 0.33–0.36 4032 1 min T (C) 11.3 11.6 3.4 2.5–18.8 37,388 1 min RH (%) 75 77 14 25–98 37,388 1 min P (hPa) 1016 1017 780 919–1032 37,388 1 min
A large variability of aerosol properties can be observed during the four weekends (WE) of the field measurements (Figure2). This variability reflects in part the different conditions above mentioned. On the second weekend only (11–12 February 2017), fairly stable weather conditions occurred: the highest weekend concentrations were measured. On the first weekend (4–5 February 2017), third weekend (18–19 February 2017), and fourth weekend (25–26 February 2017), bad weather conditions occurred, instead: we measured lower concentrations, but different aerosol properties (e.g., the AAE). Differences may be explained as follows: (i) on the third weekend only, there was no Saharan dust advection; and (ii) on the fourth weekend only, a ban on private car traffic was introduced in downtown Rome (including the CARE measurement site), and an episode of wood burning occurred.
Figure3shows (a) the wind speed distribution and (b) the integrated eBC mass concentration for different wind angular sectors.
Figure 3.Wind rose and dependence of eBC on wind. (a) Occurence rose diagram showing the joint probability density function of wind speed and direction vs. both wind speed and direction; (b) 24-h average eBC concentration vs. both wind speed and direction.
The distribution shows three major features: strong winds from E-SE, moderate winds from W-NW, and low winds which may occur from all directions (Figure3a). As expected the higher the wind speed, the lower the eBC mass concentration (Figure3cf. Figure S8 of the supplementary materials). Conversely, higher concentrations are found for lower wind speeds from all angular sectors (i.e., around the source—both dusty and non dusty days show this tendency, Figure S7 of the supplementary materials). In Figure S9 of the supplementary materials, we show a dependence of eBC on the inverse of Monin Obukhov length (z/L), whereas in Figure S25 we show the relevant time series. The z/L is the ratio of the height of the sonic anemometer sensor (z = 3 m) to the Obukhov length (L), defined as (Equation (9)):
L= u 3 ∗·Tv0 k·g· (w0T0 v) (9)
where u∗is the frictional velocity, Tv0is the mean virtual potential temperature, w0Tv0 is the Heat flux
(i.e., surface virtual potential temperature flux), k is the von Kármán constant, and g is the gravitational acceleration. The absolute value of L indicates the height to which the convective turbulence (due to the buoyancy) begins to prevail over the turbulence due to the mechanical production (mainly due to the wind shear). The sign of L depends on the sign of the heat flux (w0T0
v) only (all the other
quantities are always>0): (i) under convective conditions (heat flux>0) L is negative; (ii) under stable conditions (heat flux<0) L is positive; (iii) under the neutral condition (heat flux = 0) L = 0. The dependence between eBC and z/L (Figures S9 and S25) indicates that higher eBC concentrations occurred during stable to neutral atmospheric conditions, corresponding both to larger emissions and lower turbulent mixing.
3.2. Aerosol Characterisation
Figure4presents diurnal and weekly cycles of carbonaceous aerosol properties (mass and number concentrations, optical variables, and major sources).
Aerosol properties show clear diurnal and weekly cycles. A statistical summary of all data is given in Table2. Table3presents a statistical summary of data measured during three case-study periods: the morning rush hour of the working days, the evening peak of the working days, and on the weekends at midday. All aerosol properties are presented in more details in the following paragraphs. 3.2.1. Black and Elemental Carbon
Different measurement techniques were coupled to obtain a robust estimate of BC and EC (Section2.3.1). Data measured with different techniques show good correlations (r2>0.98, p<0.001). However, the eBC and EC mass concentration differs widely from one measurement technique to another (Table2). Figures S1–S4 of the supplementary materials show relevant results of the linear regression analysis.
The eBC mass concentration from MAAP (2.6 ± 2.5 µg · m−3) is lower than the eBC mass concentration from AE-33 (2.9±2.8 µg·m−3). Note that the mass absorption coefficient used from the manufacturer to obtain this eBC value is 7.77 m2·g−1[25], lower than the site specific MAC obtained during CARE (Section3.2.2). The linear regression analysis coefficient (i.e., the slope of the regression line) is 1.11 (Figure S1). The eBC mass concentrations from MAAP and AE-33 are both higher than the EC mass concentration from 2-h Sunset (2.1±1.4 µg·m−3), the linear regression analysis coefficient of 1.19 and 1.35, respectively (Figures S3 and S4).
Regardless of the instrument used, the eBC mass concentration shows a clear diurnal variability with higher values at the rush hours and large differences between weekdays and weekends (Figure4a—note that all Figures show eBC mass concentration from MAAP, unless otherwise specified). These differences are particularly evident at the morning rush hour (7.00–10.00 h, local time) when the mean value during the working-days is 3.9 (95% CI = 3.7–4.0) µg·m−3(Table3). The highest eBC mass concentration is measured at the evening rush hour (20.00–1.00 h), with mean value of 5.2 (95% CI = 5.0–5.5) µg·m−3. The lowest eBC mass concentration is observed during the central part of the day (14.00–16.00) of the weekends, the mean value being 1.0 µg·m−3.
3.2.2. Mass Absorption Coefficient
The Mass Absorption Coefficient of BC in the aerosol (MACBC)—a fundamental input for radiative
trasfer models—was calculated as (Equation (10)):
MACBC(λ) = σa(λ)
EC (10)
where λ = 637 nm, EC is the mass concentration from 2-h Sunset, and σais the absorption coefficient
from MAAP after the correction indicated in Equation (1). Site-specific MAC was estimated by the linear regression analysis of σa(637)and EC (r2 = 0.994, p<0.001, n = 297, Figure S10 of the
Supplementary materials). Estimated this way, the average MACBC at 637 nm was 8.7 m2 · g−1
(Std Error = 0.046). Also, MACBCat 637 nm was estimated as point-to-point ratio of σato EC (panel d
of Figure4), the mean value being 8.6 (95% CI = 8.5–8.7) m2·g−1. Larger values were observed at
the evening rush hour and during week-ends (Table3). Future research will analyse major factors govering MACBCvariability.
Importantly, the use of this site-specific MACBCvalue clearly reduces the discrepancy between
Figure 4. Diurnal and weekly cycles of: (a) eBC mass concentration; (b) OA mass concentration; (c) Mass Absorption Coefficient (MAC) at 637 nm; (d) total particle number concentration; (e) median diameter of the particle surface size distribution; (f) EC-to-OC ratios; (g) wind speed; (h) Absorption Ångström exponent (AAE) at 450–660 nm; (i) single scattering albedo (SSA) at 637 nm; (l) turbulent kinetic energy (TKE); (m) heat flux; (n) the inverse of the Monin Obukov length (z/L); (o) hydrocarbon-like OA mass concentration; (p) biomass burning OA mass concentration; and (q) oxidised OA mass concentration. The time resolution of data used to calculate the statistics (TR) is reported in Table2(if more than one TR, refer to the shortest one). Red and green bars show working days and weekend values, respectively. Data are presented as box-and-whisker plots. The boxes show median, first and third quartile, and the whiskers show the range of values that falls within the inner fences of the data (1.5·the interquartile range). Circles show outliers (outside the outer fence) and asterisks show suspected outliers (between the inner and outer fence).
Table 3.Temporal variability of major variables acquired between 2 February 16.00 and 28 February 10:00 2017 at the Rome site, separately shown for (case i) the morning rush hour (08:00 local time) of the working days, (case ii) evening peak (21:00) of the working days, and (case iii) weekends at midday (15:00). Data are presented as mean values and mean 95% confidence internal, relative number of datapoints used in the statistics (counts), and time resolution of data used to calculate the statistics (TR) being indicated.
Variables Case Mean (95 % CI) Counts TR
eBC (µg·m−3) (i) 3.9 (3.7–4.0) 962 1 min (i) 3.7 (3.2–4.2) 60 1 h (ii) 5.2 (5.0–5.5) 472 1 min (ii) 5.4 (4.7–6.0) 74 1 h (iii) 1.0 (1.0–1.0) 996 1 min (iii) 1.2 (1.1–1.3) 41 1 h MAC637 (m2·g−1) (i) 8.9 (8.6–9.2) 33 2 h (ii) 9.3 (8.2 –10.3) 8 2 h (iii) 9.1 (8.6–9.5) 16 2 h OA (µg·m−3) (ii)(i) 12.4 (10.0–14.8)10.3 (7.8–12.8) 3533 30 min30 min
(iii) 2.7 (1.6–3.7) 13 30 min NO3(µg·m−3) (i) 5.0 (3.7–6.3) 35 30 min (ii) 3.3 (2.5–3.3) 33 30 min (iii) 0.6 (0.3–0.8) 13 30 min NH4(µg·m−3) (i) 2.0 (1.5–2.4) 35 30 min (ii) 1.8 (1.4–2.3) 33 30 min (iii) 0.6 (0.7–0.9) 13 30 min SO4(µg·m−3) (i) 1.6 (1.2–1.9) 35 30 min (ii) 1.6 (1.3–1.9) 33 30 min (iii) 0.6 (0.4–0.9) 13 30 min NR-PM1(µg·m−3) (i) 19 (15–23) 35 30 min (ii) 20 (16–23) 33 30 min (iii) 4 (3–6) 13 30 min PM1(µg·m−3) (i) 21 (14–28) 15 2 h (ii) 23 (17–28) 16 2 h (iii) 5 (4–6) 8 2 h PM10(µg·m−3) (i) 29 (21–36) 15 2 h (ii) 35 (28–43) 16 2 h (iii) 14 (5–22) 8 2 h eBC/PM1 (i) 0.20 (0.16–0.24) 16 2 h (ii) 0.19 (0.15–0.22) 16 2 h (iii) 0.13 (0.11–0.16) 8 2 h eBC/PM10 (i) 0.14 (0.11–0.16) 15 2 h (ii) 0.11 (0.09–0.14) 15 2 h (iii) 0.07 (0.04–0.09) 8 2 h Ntot(cm−3) (i) 20.5 (19.3–21.6)·103 197 5 min (ii) 17.6 (16.1–19.1)·103 177 5 min (iii) 7.9 (7.0–8.8)·103 96 5 min dmed(S)(nm) (i) 80.0 (77.8–82.2) 207 5 min (i) 85.6 (84.2–87.0) 198 5 min (i) 73.8 (70.3–77.4) 96 5 min SSA 635 (i) 0.702 (0.686–0.718) 203 1 min (ii) 0.716 (0.699–0.733) 214 1 min (iii) 0.843 (0.831–0.855) 96 1 min AAE 470–660 (i) 1.32 (1.31–1.33) 1 min (ii) 1.39 (1.38–1.41) 471 1 min (iii) 1.29 (1.28–1.30) 996 1 min
3.2.3. Absorption Ångström Exponent
The Absorption Ångström Exponent (AAE) describes the wavelength (λ) dependence of the light absorption coefficient by aerosol (σa) calculated as (Equation (11)):
AAE(λ) = −dln(σa)
dln(λ) (11)
The AAE of the bulk aerosol obtained from AE-33 aethalometer (Section2.3.2) is presented in Figure2a (as data color code), daily and weekly cycles in Figure4h. The AAE of the bulk aerosol at 470–660 nm (AAE470−660) was 1.41±0.21, and varied from 0.99 and 2.03 (Table2). This indicates a
significant influence of primarily emitted BC at the site, as well as that traffic was not the only source. Lower AAE470−660occurred at the rush hours of the working days (AAE450−660of about 1.3, Table3).
Larger AAE470−660values occurred during the night (01.00–5.00, and correspond to larger values of OA
and aerosol size and lower values of the EC-to-OC ratios (Figure4b,e,f). Future research will analyse this correspondence corroborating previous results which have indicated the strong dependence of AAE on both aerosol size and composition [12,55–59].
3.2.4. Single Scattering Albedo
The Single Scattering Albedo (SSA) is a key parameter to understand the fraction of aerosol light extinction due to absorption, and thus to evaluate aerosol warming or cooling effect. SSA is calculated as (Equation (12)):
SSA(λ) = σs(λ)
σs(λ) +σa(λ) (12) The SSA of the bulk aerosol at 637 nm is presented in panel i of Figure4. This was obtained from Equation (12) at λ = 637 nm, being σs the scattering coefficient from the nephelometer and
σa the absorption coefficent from MAAP introduced after the correction indicated in Equation (1)
(Section2.3.2).
Calculated this way, the mean SSA637was 0.778 (95% CI = 0.776–0.781), Table2. The lower values
(0.702, 95% CI = 0.686–0.718) occurred at the morning rush hour (Table3). 3.2.5. Aerosol Particle Numbers and Size
Aerosol particle number size distributions (PNSDs) from 0.008 to 10 µm (electrical mobility diameter ) were calculated by combining Scanning Mobility Particle Sizer (SMPS) and Aerosol Particle Sizer (APS) data (Section2.3.13). Relevant time series measured during the CARE campaign are presented in Figure2c. The entire PNSD shows clear diurnal cycles, in particular during the central part of the campaign. The diurnal cycle is less pronounced during the two periods characterized by dust advection. Note that during these two periods no daily variability of coarse mode particles can be recognised (Figure2c). The diurnal cycle of the total particle number concentration from these data (Ntot) is presented in Figure4d. The Ntotis larger at the morning rush hour (20.5 × 103cm−3,
Table3) and evening rush hour (17.6 × 103cm−3), and lower (7.9 × 103cm−3) at midday and night time. Ultrafine particles peak at the rush hours, as clearly shown in Figure S11 of the supplementary materials presenting the number concentration and size distribution of particles from 8 to 800 nm only. Condensation mode particles (diameter from 100 to 200 nm) are recognised at night (Figure S11), as typically found in urban areas. A very few nucleation bursts occurred at midday during the week ends.
The median diameter of the entire particle surface size distribution (PSSD) from 0.008 to 10 µm (dmed) was calculated. The dmedis the electrical mobility diameter where 50% of the PSSD is lower
than dmed, and is intended to represent aerosol size relevant for particle surface area. The dmedwas
on average 81±23 nm (Table2), and has the daily variability in Figure4e. The dmedwas lower at
nocturnal rush hour. The largest dmedwere obtained at night (02:00–05:00, LT) and correspond to the
increase of the urban condensation mode particles, AAE, and OA, as mentioned before. 3.2.6. Major Non Refractory PM1Chemical Components
The time series of the non refractory PM1(NR-PM1) from the speciation analysis performed with
an Aerosol Chemical Speciation Monitor (ACSM) (i.e., organic aerosol, nitrate, sulfate, ammonium, and chloride) is presented Figure2b. Similarities among certain components are evident. The organic aerosol (OA) trend is in agreement with that of NR-PM1, since OA represents the main part of
NR-PM1. Among secondary ions, sulfate (SO2−4 ) and ammonium (NH+4), show generally similar
trends while nitrate (NO−3) shows a higher correlation with the NR-PM1. This similarity is emphasized
in Figures S12 and S13 of the supplementary materials. On average, OA accounted for 58% to the total NR-PM1, NO−3 was the second largest component (20%), NH+4 and SO2−4 both accounted about 10%,
and Cl−was about 1% (Figure S14 of the supplementary materials).
Table2summarises the descriptive statistics for these variables during the sampling period. The OA average concentration from ACSM (7.8 ± 6.8 µg · m−3) is larger than that of OC from Sunset (5.6 ±4.0 µgC· m−3), since the OC concentration only includes the carbon mass and the OA concentration also includes other elements (e.g., H and O). Figure S6 of the suplementary materials shows the linear regression analysis between OC and OA (r2 = 0.987). Also, Table 2
summarises the descriptive statistics for total particulate Polyaromatic Hydrocarbons (PAHs) during the sampling period.
An internal PMF-ME2 procedure (SoFi [60]) allows identification of emissions sources from the matrices of OA m/z measured by ACSM. Three sources were recognised: Vehicular traffic emission (HOA), Biomass burning emission (BBOA), and Oxygenated secondary aerosol (OOA). On average the primary OA (HOA + BBOA) accounted for 24% of total OA, with an equivalence between BBOA and HOA, while 72% of the total OA consisted of oxidized OA (OOA) (Figure S15 of the Supplementary materials). The unexplained OA mass represents 4%. Figure S16 of the supplementary materials present the time series of the 3 factors (BBOA, HOA, OOA) compared to the whole OA, and in Figure S17 the HOA time series is compared with that of total PAHs (30-min average)—r = 0.88, calculated on 1013 samples, which will be further analysed in future.
Figure 4o,p,q show the daily cycle of these factors for the working-days and week-ends . The bimodal trend typical of vehicular traffic emissions is clearly evident in HOA tendency (Figure4o) during the working days, with a pronounced peak at the morning rush hour (8:00) corresponding to that of eBC, and Ntot(Figure4a,b). Note that BBOA (Figure4p) shows no clear peak at 08.00 BBOA
increases during the late night (0:00–4.00) regardless of the day of week, this increase corresponding to that of AAE, dmed, OA and partly OOA. The tendency of OOA is peculiar (Figure4q) with larger
values starting from midnight until 08.00 of the day after. Future reseach will analyse this point. 3.2.7. Water-Soluble OA and Brown Carbon
In order to investigate the contribution of organic aerosol to particle optical properties, we measured light absorption of water-soluble BrC, as described in Section2.3.8. Figure5a reports the time trend of light absorption at 365 nm, together with levoglucosan ambient concentrations.
The trend of BrC absorption coefficient follows very well the levoglucosan trend, suggesting that the main source of BrC in urban Rome in winter is wood combustion. We then measured the Absorption Angstrom Exponent AAE330−500fitting the absorption logarithm as a function of the
wavelength logarithm in the range 330–500 nm. The AAE describes the wavelength dependency of light absorption (Equation (11)). Figure5b shows that AAE330−500varied between 3 and 7, in agreement
with literature values [55,61]. The AAE time trend indicates that the optical properties of water-soluble carbon changed during the experiment, with larger AAE during the last days of the month (after February 21). OC optical properties during CARE will be further investigated.
Figure 5.Time trend of levoglucosan concentration together with BrC absorption coefficient at 365 nm (panel (a)); AAE330−500time trend (panel (b)).
3.2.8. Streaker Sampler Data: Elemental Composition and Absorption Coefficients with Hourly Resolution
Elemental composition and optical data with hourly resolution were obtained by the analysis of samples collected with a streaker sampler with Particle Induced X-ray Emission (PIXE, see Section2.3.10) and multi-λpolar photometry (see Section2.3.3),respectively. The streaker sampler collects fine (PM2.5) and coarse (PM2.5−10) particles on two different substrata that, paired in a
cartridge, rotate under the sampler inlet. Therefore, particles are collected by impaction (coarse fraction) and filtration (fine fraction) in continuous streaks that may be analysed point-by-point to retrieve elemental composition and absorption coefficients at different wavelengths. Cartridge rotation and sucking orifice dimensions define time resolution to 1 h. For the aims of this campaign, analyses were focused on the fine fraction (PM2.5) and it is the first time that both elemental and absorption
coefficients detected on the same streaker frames during an ambient monitoring campaign are shown. This improvement to streaker samples analysis extends the traditional elemental characterisation to optical properties and—when suitable MACs are available—to black and brown carbon assessment without the need of additional on-site instrumentation. It is noteworthy that black and brown carbon are important tracers for PM sources so that the possibility of retrieving their concentrations together with a wide elemental characterisation allows an enhancement in source apportionment studies. Indeed, streaker data with 1 h-time resolution are widely used in the literature (e.g., [62–66] and references therein) to achieve a better identification and quantification either of sources showing a sub-daily modulation (e.g., traffic or wood burning for domestic heating) and/or impacting for a limited time slot (e.g., Saharan dust/wildfires advection events, sporadic industrial emissions or fireworks, . . . ).
The absorption coefficient measured by the polar photometer at 635 nm compared very well with σadata retrieved by MAAP (Equation (1)) being the regression slope 0.96 (±0.01), zero intercept, and
r-squared 0.98. The σa(637) from MAAP and σa(635) from the polar photometer were directly compared
as the nominal wavelengths of the two instruments (637 and 635 nm, respectively) are expected to provide about 0.3% relative difference on the measured σa. In Table2absorption coefficients measured
on streaker samples at 4 wavelengths are given (note that every time σavalues were below the detection
limit (DL) – evaluated in the range from 5 to 11.7 Mm−1as reported in [28]—half of the DL value was used).
Among the elements detected by PIXE (Section2.3.10) there are markers of several emission sources (e.g., Na and Cl for sea-salt, Al and Si for mineral dust, Cu and Zn for traffic), as well as
elements harmful for human health (e.g., Pb). In Figure6hourly data of Cu concentration (obtained by PIXE analysis) and eBC values (retrieved from σaat 635 nm on the same streaker frame) are reported
for the whole campaign. Only data with both parameters detected are represented, missing data being related to periods when the σaor Cu values were below the detection limit. Figure6shows an example
of the good correlation between Cu and eBC, which is also broadly emitted by traffic.
Figure 6.Hourly data Cu concentrations obtained by Particle Induced X-ray Emission (PIXE) analysis) and eBC values retrieved from absorption coefficients measurement by polar photometry using the site-specific Mass Absorption Coefficient (MAC) value estimated for this campaign.
The eBC mass concentrations were retrieved using the site-specific mean MAC value determined for this campaign and reported in Table2(8.6 m2·g−1). The similarity in temporal patterns suggests that the same source is likely affecting the detected concentrations and—for this specific case—it may be probably identified as traffic because eBC and Cu are well-known markers for this source (e.g., [64] and references therein). A detailed source apportionment study will be shown in a dedicated paper. 3.2.9. The EC-to-OC Ratio
The EC-to-OC ratio was calculated from the 2 h-Sunset data (PM2.5size cut-off, Table1) both
as point-to-point EC-to-OC ratio and by linear regression analysis. The temporal variation of the EC-to-OC ratio estimated as point-to-point EC-to-OC ratio is illustrated in Figure4f, mean EC-to-OC ratio being 0.38±0.22. The linear regression analysis of EC versus OC is illustrated in Figure S5 of the supplementary materials, the slope of EC vs. OC being 0.45 . The good correlation between EC and OC (r2= 0.855) indicates common emission sources (e.g., fossil fuel and biomass burning). OC can also be emitted from other (mainly biogenic) sources and form in the atmosphere after oxidation of organic compounds in the gas phase (SOA). The EC-to-OC ratio shows larger values at rush hours, indicating the dominating role of combustion sources. Lower values in the early afternoon during weekends suggests the influence of secondary organic aerosol formation. Future research will analyse its correspondence with MACBC.
3.2.10. PMxand BC to PMxRatios
Table2provides a summary of PMxmeasurements during the CARE campaign. The PM10was
26±9 µg·m−3, PM2.5= 17 µg·m−3, and PM1= 16±11 µg·m−3. These PM10and PM2.5values are
